Automatic Modulation Classification: Principles, Algorithms and Applications

5Modulation Classification Features

5.1 Introduction

In Chapters 3 and 4, two decision theoretic-based approaches to AMC are presented. For certain modulations such as AM and FM, it is obvious that signal distribution is not the only way to differentiate the two modulations. By analyzing the nature of the modulation technique, one can easily identify the key features of a signal modulated using a specific modulation scheme. While decision theoretic-based classifiers provide excellent classification accuracy, their high computational complexity motivates the development of feature-based classifiers which yield sub-optimal performance for much lower computational requirements.

In this chapter we list some of the well-recognized features designed for modulation classification. While the book focuses on digital modulations, features historically used for the classification of analogue modulations will also be included in this chapter. In the remainder of the chapter we first investigate the spectral-based features which exploits the spectral properties of different signal components. The wavelet-based features are given as another approach to feature-based modulation classification using the signal waveform. The high-order statistic features are examined as optioned to classifier digital modulations of different type and orders. The cyclic features based on cyclostationary analysis are presented at the end. More advanced feature-based (FB) classifiers which could utilize the features ...